Watch on YouTube
(Embed)
28
0
Genre: Science & Technology
License: Standard YouTube license
Family friendly? Yes
Shared September 28, 2026
Three Indian bug hunters from Hackron AI hacked OpenAI in 72 hours using Anthropic's Claude. They exploited a crafted iPhone HIC image file to trigger a vulnerability in OpenAI's forum server image processing library. Then they leveraged a single sign-on flaw to access an employee's Codex account and private GitHub repository. The team spent $3,000 and received $6,500 from OpenAI's bug bounty. Claude assisted throughout—investigating vulnerabilities, writing exploit code, and debugging in real time. This breach highlights AI's dual-use risks and the urgent need for stronger cybersecurity in AI labs.
